vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge branch 'master' into gg/flash-attn

Browse source
This commit is contained in:
Georgi Gerganov 2024-03-04 20:42:48 +02:00
commit e307882c34
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
17 changed files with 995 additions and 272 deletions
Download patch file Download diff file Expand all files Collapse all files

17
ggml.h
View file

@ -454,6 +454,8 @@ extern "C" {
GGML_OP_POOL_2D,
GGML_OP_UPSCALE, // nearest interpolate
GGML_OP_PAD,
GGML_OP_ARANGE,
GGML_OP_TIMESTEP_EMBEDDING,
GGML_OP_ARGSORT,
GGML_OP_LEAKY_RELU,
@ -1662,6 +1664,15 @@ extern "C" {
int p2,
int p3);
// Ref: https://github.com/CompVis/stable-diffusion/blob/main/ldm/modules/diffusionmodules/util.py#L151
// timesteps: [N,]
// return: [N, dim]
GGML_API struct ggml_tensor * ggml_timestep_embedding(
struct ggml_context * ctx,
struct ggml_tensor * timesteps,
int dim,
int max_period);
// sort rows
enum ggml_sort_order {
GGML_SORT_ORDER_ASC,
@ -1673,6 +1684,12 @@ extern "C" {
struct ggml_tensor * a,
enum ggml_sort_order order);
GGML_API struct ggml_tensor * ggml_arange(
struct ggml_context * ctx,
float start,
float stop,
float step);
// top k elements per row
GGML_API struct ggml_tensor * ggml_top_k(
struct ggml_context * ctx,